Superconducting Nb3Sn joints by chemical vapour deposition

Superconducting Nb3Sn joints by chemical vapour deposition

Superconducting Nb3Snjoints by chemical vapour deposition P.G. Kosky, H.C. Peters, C.L. Spiro, D.S. McAtee, L. Rumaner* and D. Marsh GE Research and D...

750KB Sizes 4 Downloads 49 Views

Superconducting Nb3Snjoints by chemical vapour deposition P.G. Kosky, H.C. Peters, C.L. Spiro, D.S. McAtee, L. Rumaner* and D. Marsh GE Research and Development Center, PO Box 8, Schenectady, NY 12301, USA *University of Washington, Seattle, WA 98195, USA Received 13 October 1993; revised 18March 1994 Nb3Sn is an A15 superconductor with high critical current density in high magnetic fields, e.g. 200000Acm -2 at 10T and 4.2 K. Unfortunately, A15 intermetallics are brittle and difficult to manufacture in defect-free lengths for large superconducting coils. This paper describes a chemical vapour deposition (CVD) process which deposits Nb3Sn on to prespliced Nb3Sn substrates to form a superconducting joint between two superconductive sections. The CVD-deposited material has very fine grain structure (probably less than 0.2/zm), a Tc of between 16.2 and 17.8K (in zero field) and a critical current density of greater than 500000Acm -2 at 5T. In addition, the CVD layer definitively bridged the gap between two spliced superconductors, increasing the current capacity at 5T from a baseline capacity of 66Amm -1 on the original material to greater than 80Amm -1 after a CVD Nb3Sn coating of 3.5/~m per side had been applied.

Keywords: Nb3Sn; superconducting joints; chemical vapour deposition

The intermetallic compound Nb3Sn is a type II superconductor with a high critical current density in high magnetic fields, e.g. 200000Acm -2 at 10T and 4.2K. It can be synthesized by the reaction of liquid tin on solid niobium 1. Its synthesis by this method is possible because NbaSn is thermodynamically stable 2'3 at reaction conditions. A typical cross-section of a finished multilayer product is given in Figure 1. The central Nb core is what remains of the original niobium; the Nb3Sn layers, which are formed by diffusion of tin, are soldered to a mechanically and thermally protective copper cladding. The copper provides some integrity against the extreme brittle behaviour of the Nb3Sn and is a precaution against superconductive quenching. For the production of very large windings several miles of defect-free superconductor often are required. Unfortunately periodic defects are inevitable and a means of splicing lengths of the superconductor is necessary4'5. A non-superconducting joint in a superconducting winding is satisfactory from the thermal point of view if the local Joule heat generated does not cause the remainder of the winding to quench to its normal state. However, for long-term magnetic field stability, a winding should be operated without any dissipative regions and so avoid the complication of a continuous external power supply. A conceptually 0011-2275•94•090753-07 (~ 1994 Butterworth-HeinemannLtd

simple solution is to use a superconducting solder to join two pieces of good quality superconductor. Unfortunately, common solders have poor critical current densities and/or low critical magnetic fields. A preferable method therefore envisages the use of high field, high critical current superconductors as a specialized 'solder' superconductive junction between discrete superconductive sections. Nb3Sn does not form a thermodynamically stable phase below 920°C; however, it has been made 6 by non-equilibrium chemical vapour deposition (CVD) methods at temperatures ranging from 730 to 1600°C. Other niobium superconductors also have been produced by CVD methods. For example Nb3Ge, an intermetallic with high Jc, He2 and Tc, has a relatively narrow thermodynamic range 7'8 but has been produced at non-equilibrium temperatures 9"1°. For CVD coating of Nb3Sn, a modified Van Arkel Nb3Sn layer


Sn/Solder 91 / Nb3Sn layer

Nb c~Ore Copper Cladding

Figure 1 Cross-section of as-received ribbon (not to scale)

Cryogenics 1994 Volume 34, Number 9 753

Superconducting Nb3Snjoints by chemical vapour deposition: P.G. Kosky et al. reaction can be used (1)

3NbCI5 + SnCI4 + 9.5H2 = Nb3Sn + 19HC1

It should be emphasized that this is not an elementary reaction; what is written is merely the ideal net result. A relatively simple apparatus is necessary to demonstate the deposition of an A15 Nb3X phase on a short sample of substrate; however, the sealing requirements are critical since niobium has a high reactivity to oxygen and yields the oxides NbO2 (AfG ° = -739.2 kJ/g.atom Nb) and Nb205 (AfG ° = -882.9kJ/g.atom Nb) which are comparable in stability to ot-A1203 (Af G ° = - 837.9 kJ/g. atom AI). During this study, oxidation of Nb-rich materials was observed even in the U H V chamber of an Auger spectrometer and it may be correctly concluded that oxidation at atmospheric pressure is rapid. For this reason the CVD system must be well sealed. We purged our system with deoxygenated and dried helium gas to a residual level of 50ppb of 02 and H20, respectively.

Apparatus A schematic of the apparatus is shown in Figure 2. Gas flows were monitored using mass flow controllers and reactor temperatures were monitored by thermocouples. The source gas for Nb was generated at about 375°C by passing C12(g) in a He carrier gas over chips of Nb producing volatile chlorides by direct chlorination. A 25 mm diameter quartz tube was filled with niobium chips and placed in the chlorination furnace. Hanak et al.6 used high temperature (800°C) chlorination and claimed NbC14(g) as their product. Asano et al. 9 worked at 350°C and claimed NbCls(g). Brisbin and Heffernan 11 worked at 340°C and found that the amount of Cl2(g) consumed corresponded to 20-30% less than was necessary to yield NbCIs; they suspected the presence of NbC13 and NbC14 in the nominal NbCIs. Other known compounds 12 of niobium and chlorine include NbC12 and NbClx (x = 2.67 and 3.13). Evidently one should have some concern as to the exact nature of the chloride produced. While, in H2

principle, thermodynamics should distinguish between the possibilities and predict the favoured species at a given condition, non-equilibrium effects such as mass transfer (see Appendix) always leave some doubt as to the efficacy for the chlorination of niobium. Apart from the question of stoichiometry, one problem is that NbC13 is only marginally volatile below about 700°C. While the Nb chlorides are made by passing C12(g) over chips of large surface area, tin melts at 232°C and the subsequent planar surface area presented to the chlorine is much reduced compared to that of solid chips, which would thus inhibit mass transfer. Tin forms two stable volatile chlorides, SnC12 and SnC14. Hanak et aL 6 produced SnCI2 (g) at 800°C but SnCI4 is obviously more preferable because of its enhanced volatility at lower temperatures. Ultimately it was easier to use an evaporator containing SnCI4 (1) and to evaporate it in a stream of helium gas. Standard vapour pressure data were used 13 to calculate the concentration of vaporized SnCI4. Generally the thermostat bath containing the SnCI4 (1) evaporator was operated at at temperature corresponding to a vapour pressure of about 48 torr*. Later we approximately doubled the concentration of the flowing halides relative to inert gas by increasing the SnCI4 reservoir vapour pressure to 78 torr. As shown in Figure 2, the mixed gases flowed into a high temperature furnace (700-950°C), and there mixed with H2 to promote reduction reactions in the furnace. The H2 flowed concentrically in an annular space surrounding the tube containing the mixture of flowing gaseous halides so that t l ~ latter was kept sufficiently cool to prevent pyrolysis and subsequent deposition in that location. The H2 and the halides were mixed inside the furnace just ahead of the sample position. The optimum position relative to the samples was found empirically. The reactor was about 45cm long and 5cm in diameter and was heated over its middle 30 cm. The test coupons, of length L (approximately 5-10cm), were located empirically in the hot zone at a location which gave 100% coverage of deposit. Once the 1-I2and the halides mixed in the hot section of the reactor, reduction occurred to Nb, Sn and their alloys, as well as some side reactions with minor components.

CVD Reactor

Experimental procedures Nb chips p~

1 Sample Coupon

° l

He Effluent CI2/He

-",-- SnCI4 saturator

Figure 2 Schematic of CVD apparatus for deposition of A15 phase of NbsSn

754 Cryogenics 1994 Volume 34, Number 9

The following substrates were used: HastelloyT M X, sapphire, a-alumina, Nb and Nb3Sn. These substrates were either single strips for analysis purposes or joints formed from two pieces of substrate. Test specimens were cut into pieces approximately 10 cm long. A grey Nb oxide layer was present on any air-exposed surface of Nb3Sn or Nb coupons as confirmed by an Auger scan.

"1 torr = 133.322 N m -2

Superconducting Nb3Snjoints by chemical vapour deposition: P.G. Koslo/ et al. Table 1 Typical reactor conditions


Experiment 1 CI2(g) to chlorinator


Experiment 2 36.1

(sccm) a NbCI5(g) to reactor at 100% conversion (sccm) Helium carrier gas to chlorinator (sccm) He/CI2(g) Temperature of SnCla reservoir (°C) Vapour pressure of SnCI4(v) (torr) Helium flow to SnCI4 reservoir (sccm) Mole fraction SnCla(v) in helium SnCla(v) flow to reactor









=s o





I' L I I



16.3 37

8.3 49





I /








15.3 1:2.25 101.3

7.6 1:2.27 101.5

470 4.64

500 4.93












] 20






I 25







Figure 3 Superconducting transition temperature in a CVD layer of Nb3Sn on (x-AI203

(sccm) He/SnCI4(v) Nb/Sn in reactor Stoichiometric sccm of H2 Actual sccm of H2 H2 actual/ H2 stoichiometric Reactor temperature (°C)


SCCM = standard (0°C, 1 bar) cm s rain -1

scan. The substrate was weighed and placed on a ceramic boat in the sample furnace. The chlorination furnace and the SnCI, transfer line were purged with helium gas throughout the test. The chlorination furnace was set at 375°C and the sample furnace was set to 835°C. When the sample furnace reached its assigned temperature, chlorine gas was conducted to the niobium chips and helium was redirected through the SnC14 vesel. Finally hydrogen was admitted to the sample furnace. A summary of some successful reactor conditions is given in Table 1. The deposition usually lasted for 20min, although the full furnace cycle was nearer to 2h. At the end of the deposition, the chlorine flow to the chlorination furnace, the helium flow through the SnC14 bath, and the H2 flow to the CVD reactor were turned off and the samples cooled in flowing helium. The substrate was removed and a final weight taken. We were very concerned with the oxidation state of the substrate since NbOz/NbzO5 is easily formed as a ubiquitous black powdery residue. However, if we maintained H 2 over the substrate until the system was cool, the result was another very friable substrate containing NbHx. Nb/NbH has an eutectoidal temperature of 171°C14 and is easily formed; thus it is imperative not to cool niobium-rich materials in H2. In our successful experiments, the key was to control the oxidation state and this was accomplished eventually by sequencing the flow control valves, admitting reactive gases so the test section never saw Hz except at reaction temperature. This avoided both oxides and hydrides of niobium.

of phase composition. A diffraction pattern of an unequivocal A15 Nb3Sn deposit on a substrate of aA1203 was free of detectable impurities. The Tc of this particular sample was 16.1 K at 100 mA current (Figure 3) in the absence of an imposed magnetic field, while similar material deposited on a sapphire crystal face gave 17.8 K. Superconducting critical current measurements were made on metallic substrates at 5 T, 4.2 K after coating them with electro-deposited nickel for soldering of current/voltage leads. Critical current was measured at a ramp rate of 5 A s--1 and our criterion for quenching was a measured sample potential difference of 0.2#V that corresponded to a resistivity of about 2 x 10-13 Q cm. The critical current density jc A cm-2 is related to the measured critical current Ic by jc = 105 Ic/2a A cm -2


Table 2 Critical currents on CVD A15 NbzSn Substrate

,~ (/~m)

/c at 5T ( A m m -1)

Baseline NbzSn Coated NbaSn test coupon Incremental Nb3Sn test coupon Coated Nb3Sn test coupon Incremental Nb3Sn test coupon Coated NbzSn joint Incremental NbzSn joint Coated Nb3Sn joint Incremental Nb3Sn joint

5.6 a 13.5

66.3+10.4 b 159





Jc at 5T (Acm -2) 592000 589000 585 000






>80.0 < 100 >13.7 <33.7 80.6

>437 000 <546 000 >193000 <475 000 420 000



179 000



Characterization of samples X-ray diffraction was used as the primary determinant

a Nominal b+lo-

Cryogenics 1994 Volume 34, Number 9 755

Superconducting Nb3Snjoints by chemical vapour deposition: P.G. Kosky et where h is the thickness of the superconductor in/zm and Ic is in A mm -z. Equation (2) assumes both faces of the test coupon are coated and participating. We quote the current per unit width of superconducting since we saw only proportional changes in current by nearly doubling the width. Obviously there is some coupon width at which edge defects become important; however, our width-to-superconductor ratio is >100 and our assumptions behind Equation (2) seem justified. Table 2 gives some representative data for CVD deposited Nb3Sn and for the baseline material. The baseline material has some variation in thickness of the diffusion-deposited NbaSn layer. There is also some corresponding variability in the tested critical current between samples from the same batch of test coupons due to thickness variations or intrinsic variation in the properties of the Nb3Sn layer itself. The values of the baseline material are therefore given to one standard deviation. The set of entries after the baseline entry is for test coupons which had been CVD-coated with Nb3Sn. The total thickness (/zm) of Nb3Sn per side is the sum of the baseline and incremental layer thicknesses. For joints, two coupons were TIG welded for about 11.5 cm along their periphery to provide as intimate a contact as possible between the test coupons at the edge location (Figure 4). Not surprisingly, the coated materials outperformed the baseline material by virtue of the additional layer of superconductor, i.e. all coated samples had critical currents greater than the baseline value of 66.3 + 10.4Amm -1 at 5 T. The incremental current capacity indicates the performance of the CVD layer. This is arrived at by subtracting the mean performance of the baseline Nb3Sn. The incremental superconductive current capacity of the test coupons is approximately linear with the thickness of the CVD layer and the incremental current density is usually similar to that in the baseline material, with statistical uncertainties due to the variability referred to above. We have compared the CVD-coated joints to the original baseline material giving full credit to the current capacity of the baseline material. For example, in the Table 2 entry where Ic = 80.6 A mm -1, the NbaSn coating by CVD was 4.0/zm. Nevertheless we have included the thickness of the baseline material (5.6/zm) in the calculation ofjc even though it possibly was not participating. This is parallel to the treatment





\ ~\\\\\~\\\\\\\\\\\\\\\\~J//////////////~///////A riG WELD

Figure 4 Schematic of (a) plan view of edge TIG welded joints and (b) enlarged cross-section showing conductive path


Cryogenics 1994 Volume 34, Number 9



Sn-- ll -



CVD __~ ~ _ Nb3Sn

_ ~ 14._ CVD Nb3Sn

Figure 5 (a) SEM of as-received Nb3Sn coupon viewed across its edge and consisting of a 9 # m niobium core and two layers of Nb3Sn each of 6/~m thickness. The individual granules of Nb3Sn are between 0.5 and l # m in their major dimension. Gross particles seen external to the NbzSn layer are of residual tin. (b) SEM of reacted Nb3Sn coupon with a CVD layer of NbzSn. The niobium core has shrunk to 7.5#m and the original diffusion layer of NbzSn has grown to 8.8#m per side (for a total original coupon thickness of 25#m). The outermost layer is the CVD deposit which has an average thickness of 2/~m per side. The reaction conditions were S n : N b = 2 : l at 838°C for about 20 min

of critical current densities in the test coupons that had an unbroken base of superconductor. In essence we are assuming that the CVD coating repaired the base layer as well as providing an intimate overlayer of superconductor. The alternative view is to assume that all of the current is shunted through the incremental CVD deposit and thus its current density is (5.6+4.0)/ 4.0 = 2.40 of that quoted (i.e. 2.40×420000 or 1.01 x 106A cm-2). However, in our view, it is improbable that we can exceed the incremental critical current

Superconducting Nb3Snjoints by chemical vapour deposition: P.G. Kosky et al. density of the integral test coupons, which from Table 2 appears to be about 600 000 A cm-2. In any case, in a pragmatic sense, we only need to exceed the current carrying capacity of the baseline material to ensure that quenching will occur remote from the CVD-coated joints. A large number of specimens were used to determine the compatibility of the deposited Nb3Sn with the substrate material. The CVD-coated material was etched (using a dilute mixture of hydrofluoric, nitric and lactic acid for a few seconds) for SEM and Auger analysis. Figure 5a shows details of the cross-section of a virgin test coupon and Figure 5b shows the same after reaction. Two important conclusions are that the Nb core has been depleted compared to the original with concomitant growth of the Nb3Sn layer. The original Nb3Sn is distinguishable from the CVD layer by its large grain size. The CVD layer of Nb3Sn is the dearly visible layer of about 2/zm per side (in this case). The bonding on the CVD layer is intimate with the substrate, lending some credence to the thought that nucleation had occurred at the interface, which should further ensure that the deposited overlayer has the A15 structure. Figure 6 shows just how intimate the bonding is to the base Nb3Sn layer. The grain of the latter is dearly visible, but the CVD layer shows no detectable grains. Finally we checked with Auger spectroscopy for the state of the interface between the CVD layer and the base Nb3Sn material. Figure 7 shows that the CVD layer does not trap oxide at the interface between it and the base Nb3Sn material, although oxygen is clearly present at the outer edge of the test coupon due to atmospheric exposure. Indeed, if a non-superconducting oxide interlayer were present between the two layers of Nb3Sn materials at a

Figure 6 SEM scan of interface region between the CVD Nb3Sn layer and the original Nb3Sn diffusion layer, The former is the substantially homogeneous overlayer and the latter the granulated material directly below it. Note the apparent fineness of the grain size of the CVD layer compared to the original diffusion-deposited Nb3Sn layer









Sn -'"




j¢, Nb






Figure 7 Auger scan of elemental composition across edge of Nb3Sn coupon. The Auger sputter profile is taken in a direction perpendicular to the edge of the coupon and shows oxygen at the exposed outer surface of the CVD layer due to air contamination; but no excess oxygen is detectable at the interface of the CVD coating and the original deposit of NbzSn

thickness greater than the Cooper pair cohesive distance (of perhaps 30-50/~*), superconducting current would not have flowed.

Kinetics The rate of deposition of the Nb3Sn is dependent on several parameters. However the gas concentrations were not found to be very important in this study: thus we increased the halide concentrations by approximately a factor of two (Table 1) without an obvious increase in the deposition rate. Nor was the concentration of H2 very important, at least in the range we worked. The H2 was always in stoichiometric excess by a factor of approximately five or more; additional H2 did not materially affect the rate of deposition. The principal determinant of rate was temperature, presumably through the net effect of at least three temperature dependent processes. These are: 1, vapour phase deposition of Nb3Sn; 2, conversion of the Nb ribbon core to Nb3Sn by diffusion of a tin-rich phase through the existing NbaSn coating; and 3, the etching of Nb at temperatures below about 775°C. These can be represented by R1, R2, and R3molcm-2s -1, respectively. The last effect is evident in the data presented in Figure 8a. The overall rate is negative at the lower temperatures as the etching rate there is faster than the growth rate. Etching is probably due to the reaction of HC1 formed either as a by-product in the Van Arkel reaction [Equation (1)] or via halogen in the reactant gases which would have converted to HCI in the furnace (see the Appendix). The net rate of deposition is measured by weight change of the substrate. Then, using the density of the deposit and of the substrate material, it is easy to * I A = 10-1°m

Cryogenics 1994 Volume 34, Number 9 757

Superconducting Nb3Snjoints by chemica/ vapour deposition: P.G. Kosky et " 0.1 -~ 0 -0.1" -0,2-0.3-0.4" -0.5-









• ~



concentrations of HCI(g) and other constituents not well measured in these experiments. Therefore no attempt has been made to further correlate the kinetic parameters implicit in Equation (4). However, the net result of Equation (4) is the data shown in Figure 8a. The unweighted least squares regression line through those data is




d,t/dt = 3.12 x 10-3 T - 2.40/zm s-1





2 1 0 -1


I1. w a





where T is in °C. The principal kinetics of interest are those which determine the deposition of NbaSn compared to the Nb core consumed. That is the ratio of dh/dt to R2, i.e. Equation (6)/Equation (5). The result is shown in Figure 8b which shows an optimum temperature of about 820°C (the range 825-850°C had already been determined by empirical observation). It is important to recognize that the deposition data are inherently skewed due to C12(g) which by-passed the Nb chips in the chlorinator. An optimum might well lie at lower reaction temperatures if the etching is suppressed by improved mass transfer in the chlorination furnace or the use of pure NbC15 feed.

z -5"













Figure 8 (a) Overall rate of deposition of NbaSn and Nb core depletion as a function of temperature. (b) Relative deposition rate of Nb3Sn compared to Nb core consumption

convert the weight change into the thickness of deposit. For a substrate of density Ps and a deposit of density p, the average thickness is

,~ = ~mpsl2pm


for a sample of mass m of Nb3Sn which gained mass t~m. This expression presumes that the substrate is coated with an average coating of h per side. The room temperature densities of Nb3Sn 15 and Nb 16 are essentially equal, at 8.57 and 8.60 gcm -3, respectively. The net deposition rate is defined by the gain in the thickness of the Nb3Sn coating, which may be written in terms of its three components as



104 × M/p(Rz + R2

- -


S- 1


where M is the molecular mass of NbaSn (397.4 g mol-1). Data 17 for the reaction of Sn and Nb in the range 1000-1400°C yield R 2 = 6.32 exp (-44180/RT)

g. atom Nb consumed c m -2 s -1

Acknowledgements (5)

where R = 1.987calmo1-1K -1 and T is in K. If the reaction mechanism is unchanged at the lower temperatures of this study, this estimate can be applied to the current problem. We know of no extant data for either R1 o r R3, in any case they will depend on the

758 Cryogenics 1994 Volume 34, Number 9

The CVD technology for the manufacture of Nb3Sn superconductor is well-established; as shown here it may be modified to join two superconductive test coupons which will have a high resulting superconductive current density and a high magnetic field capability. While not described in the text, the final joint can be made mechanically comparable to the preexisting material. Additionally, the CVD layer of Nb3Sn is several microns thick and has a fine grain structure and high incremental current capacity. The baseline critical current of the substrate material was exceeded in all cases and this is sufficient for most applications, assuming the baseline material would quench before the joint. While Nb3Sn is an excellent superconducting 'solder', metallurgical challenges in the CVD process are derived from the need to sustain a particular oxidation state of the niobium and to avoid the twin pifalls of oxidation and hydride formation. Reaction conditions and stoichiometry have been chosen to achieve these things and, with the present chlorination method for the production of NbCIs, they achieve a maximum efficiency around 825-850°C.

Dr M.G. Benz assisted early in this work. Dr J.E. Tkaczyk measured electrical contact resistances and both he and Dr F.E. Luborsky measured Tc on many samples. M. McConnell performed the Auger measurements. Finally Dr B. Karas provided timely and incisive advice on electroplating and its pitfalls.

Superconducting Nb3Snjoints by chemical vapour deposition: P.G. Kosky et al. References 1 Hail, E., Benz, M.G., Rumaner, L.E. and Jones, K.D. Interface structure, grain morphology, and kinetics of growth of the superconducting intermetallic compound Nb3Sn, GE CRD Technical Report 90CRD265, Schenectady, NY, USA (1990) 2 Massalski, T.B. (Ed) Binary Alloy Phase Diagrams ASM, Metals Park, OH, USA (1986) 1698 3 Shmsk, F.A. Constitution of Binary Alloys (Second Supplement) McGraw-Hill Book Co, NY, USA (1985) 203 4 Kosky, P.G., Peters, H.C., McAtee, D.S. and Spiro, C.L. Method for producing superconductive joints US Patent 5 215 242 (1993) 5 Benz, M.G., Knudsen, B.A., Rumaner, L.E., Zabala, R.J. et al. Melt-formed superconductive joints of Nb3Sn tape, GE CRD Technical Report 93CRDl19, Schenectady, NY, USA (1993) [see also US Patents 5109583 and 5134040 (1992)] 6 Hanak, J.d., Strafer, K. and Cullen, G.W. Preparation and properties of vapor-deposited niobium stannide RCA Review (1964) XXV 342 7 Shunk, F.A., Constitution of Binary Alloys (Second Supplement) McGraw-Hill Book Co, NY, USA (1985) 184 8 Massalski, T.B. (Ed) Binary Alloy Phase Diagrams ASM, Metals Park, OH, USA (1986) 1228 9 Asano, T., Tanaka, Y. and Tachikawa, K. Effects of deposition parameters on the synthesis of Nb3Ge in the CVD process Cryogenics (1985) 25 503; and Effects of H2/C12 ratio on CVD synthesis of superconducting Nb3Ge tapes Cryogenics (1987) 27 386 10 Suzuki, M. and Anayama, T. High field properties of niobiumgermanium (Nb3Ge) and niobium nitride (NbN) films Sci Rep Res lnst Tohoku Univ Ser A (1992) 37(1) 35 11 Brishin, P.H. and Heffernan, W.J. Preliminary studies for the deposition of niobium by vapor reaction, GE Report 62GL169, Schenectady, NY, USA (1982) 12 Barin, I. and Knacke, O. Thermodynamic Properties of Inorganic Substances Springer-Verlag, Berlin, Germany (1973) 555556 13 Weast, R.C. (Ed) Handbook of Chemistry and Physics 68th Edn, CRC Press Inc., Boca Raton, FL, USA (1987) D-195 14 Shunk, F.A. Constitution of Binary Alloys (Second Supplement) McGraw-Hill Book Co, NY, USA (1985) 184 15 Pearson, W.B. A Handbook of Lattice Spacings and Structure of Metals and Alloys Vol 1, Pergamon Press, NY, USA (1958) 16 Weast, R.C. (Ed) Handbook of Chemistry and Physics 68th Edn, CRC Press Inc., Boca Raton, FL, USA (1987) B-111 17 Benz, M.G. Kinetics of condensed phase Nb/Sn reaction, personal communication, GE Research and Development Center, Schenectady, NY, USA (1990)

Appendix: Reactor conditions The reactor conditions given by Hanak et ai. 6 for the deposition of Nb3Sn suggest that the correct vapour

composition of Nb:Sn in the reactor is not 3:1 but closer to 1:4. However their recommended composition did not yield the expected product in this investigation. We were successful when the vapour phase composition was about Nb:Sn = 1:2. Table I shows an overall view of our successful range of experimental conditions. The dilution of reactive gases initially was based upon the conditions advocated by Hanak et al. ; eventually we modified these by increasing the concentration of reactive gases relative to inert gases from about 1:16 to about 1:8, but without effect on rate. Nor did increasing the H2 rate from stoichiometric H2 to an excess 10:1 change the rate of deposition of Nb/Sn. It is also interesting briefly to consider the fluid mechanics of the gases in the reactor. The essential geometry is in the sketch in Figure 2. The flow is laminar with N R e ~ 2 . 5 (based on furnace diameter) and, at this condition, there is too little convection to provide radial mixing. Axial mixing is due to convection and to diffusion. The ratio of these defines the P6clet number, Nee = V L / D ~ 10, where L is the length of the substrate in the flow direction and D is the diffusivity of the gas mixture in the reactor. Molecular diffusion did not guarantee uniform composition along the reactor. Indeed we saw the effects of nonuniformities by placing a long substrate in the reactor and this meant trial and error in determining the optimum axial location for the sample. The conditions in the chlorinator were rather similar tothose in the reactor except the flow area was smaller (25 mm packed diameter in the chlorinator compared to 50 mm open diameter in the CVD reactor); therefore the same criticisms of its operations pertain. The Reynolds number in the chlorinator was of the order of 10; NRe~>4000-10000 are needed for turbulent flow and good mixing. In particular, some CI2(g) bypassed the niobium bed and entered the CVD reactor. This is evidenced by the amount of niobium consumed in the chlorinator compared to the amount of feed chlorine; over a total run time of more than a 1000 min, the ChNb ratio was 6.61 compared to a theoretical maximum of 5. Since the experimental error in this ratio was no more than +0.4, the difference between these measured and theoretical values is meaningful.

Cryogenics 1994 Volume 34, Number 9 759